cs.CEMar 17, 2026

Generative AI for Quantum Circuits and Quantum Code: A Technical Review and Taxonomy

Authors: Juhani Merilehto

Organizations: University of Vaasa & University of Turku

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

We review thirteen generative systems and five supporting datasets for quantum circuit and quantum code generation, identified through a structured scoping review of Hugging Face, arXiv, and provenance tracing (January-February 2026). We organize the field along two axes: artifact type (Qiskit code, OpenQASM programs, circuit graphs); crossed with training regime (supervised fine-tuning, verifier-in-the-loop RL, diffusion/graph generation, agentic optimization); and systematically apply a three-layer evaluation framework covering syntactic validity, semantic correctness, and hardware executability. The central finding is that while all reviewed systems address syntax and most address semantics to some degree, none reports end-to-end evaluation on quantum hardware (Layer 3b), leaving a significant gap between generated circuits and practical deployment. Scope note: quantum code refers throughout to quantum program artifacts (QASM, Qiskit); we do not cover generation of quantum error-correcting codes (QEC).

Figures & tables

YearSystem / DatasetTypeKey Innovation
2020QASMBench [ 10 , 11 ]BenchmarkCurated low-level OpenQASM 2.0 benchmark suite for NISQ evaluation
2024genQC [ 12 , 13 ]ModelFirst diffusion model for quantum circuit synthesis; text-conditioned denoising over discrete gate tokens
2024KetGPT [ 14 , 15 ]ModelGPT-based transformer for generating realistic OpenQASM 2.0 circuits (dataset augmentation)
2024AltGraph [ 16 ]ModelGenerative graph models (D-VAE, DeepGMG) for circuit DAG rewriting and optimization
2024QiskitHumanEval [ 6 ]BenchmarkUnit-test benchmark (101 tasks) for Qiskit code generation
2024quantum-circuits-8k [ 17 ]DatasetSynthetic text →\rightarrow QASM 2.0 pairs with paraphrase augmentation
Table 1: Chronological milestones in generative AI for quantum circuits and code. Years reflect first public appearance (preprint, model card, or repository).
SystemFamilyOut-   
putSizeSyn.Sem.HWEvaluation Notes
QuantumGPT-     
124M     
[ 32 , 17 ]Spec.    
small LMQASM    
2.0124 M✓Lim—Parser validation; manual inspection on ≤\leq 5 qubit circuits (no oracle)
Table 2: Reviewed generative systems with artifact types, training regimes, and evaluation coverage.
DatasetPrimary useScaleNotes
quantum-circuits-8k [ 17 ]Text →\rightarrow OpenQASM 2.0 SFT∼\sim 8 kSynthetic with paraphrase augmentation; small-circuit emphasis
graph-data-quantum-rl [ 31 ]Optimization-circuit generation and RL14.5 k rowsPrompts, graphs, Hamiltonians, OpenQASM 3.0 circuits, solutions
QASMBench [ 11 , 10 ]OpenQASM-2 benchmark suitediverseCurated benchmark circuits and circuit-level metrics
QCircuitBench [ 20 ]Algorithm design benchmarking120,290QASM 3.0 + code (Qiskit/Cirq) + oracles / verification functions
QuantumLLMInstruct [ 18 , 19 ]Broad quantum instruction data500k+ claimedPaper/model card claim 500k+ instruction-tuning pairs across 90+ quantum domains; current public HF viewer exposes 5.15k rows
Table 3: Datasets and benchmarks supporting quantum circuit/code generation and evaluation.
Task ObjectiveRecommended EvaluatorKnown Failure Modes
Compilation to target unitaryProcess fidelity, diamond norm proxy, or equivalence checking on a basis subsetRelative phase errors invisible to basis-restricted measurement; partial basis checking misses errors on untested inputs; ancilla garbage passes fidelity but fails full equivalence
Optimization ansatz generationEnergy expectation value ++ convergence speed ++ robustness under re-optimizationLow-energy ansatz may be a local minimum; convergence speed conflated with initial parameter sensitivity
Algorithm design tasksOracle-based functional checks (as in QCircuitBench [ 20 ] )Oracle leakage if test structure correlates with training data; hard to verify beyond provided oracles
Code assistants (Qiskit)Unit tests ++ execution traces (QiskitHumanEval [ 6 ] )Tests check observable behaviour, not internal correctness; ancilla state and relative phase may be ignored
Dataset augmentationReal-vs-random classification ++ structural analysis (KetGPT [ 14 ] )Distribution matching without semantic grounding; generated circuits may be syntactically realistic but computationally trivial
Table 4: Mapping task objectives to recommended semantic evaluators and known failure modes.
More in this topic

Quantum Circuit Synthesis

See all 19 papers

Explore similar work

CardsList

  1. Quantum Circuit Vision: Cost-Aware Evaluation of Visual AI Agents for Quantum Code Generation

    Jul 11, 2026Dongping Liu, Aoyu Zhang, Luyao ZhangAI Agent EvaluationAI Agent Benchmarks

    Amazon Web Services Hong Kong, China · Amazon Web Services Beijing, China · Digital Innovation Research Center and Social Science Division, Duke Kunshan University Address: Duke Avenue No.8, Kunshan, Suzhou, Jiangsu, China, 215316

  2. QuanBench+: A Unified Multi-Framework Benchmark for LLM-Based Quantum Code Generation

    Mar 25, 2026Ali Slim, Haydar Hamieh, Jawad Kotaich +5Code Generation BenchmarksCode Generation Evaluation

    American University of Beirut · King Abdullah University of Science and Technology

  3. QTestBench: A Test Framework for Evaluating LLM-Generated Quantum Code

    Oct 9, 2026Meriem Terki, Nouhaila Innan, Minghao Shao +3Automated Test GenerationLLM Evaluation

    Ecole Supérieure en Informatique de Sidi Bel Abbès, Algeria · eBRAIN Lab, Division of Engineering, New York University Abu Dhabi (NYUAD), Abu Dhabi, UAE · Center for Quantum and Topological Systems (CQTS), NYUAD Research Institute, NYUAD, Abu Dhabi, UAE +1